Mechanistic Modeling and Multiscale Applications for Precision Medicine: Theory and Practice

Recent progress in artificial intelligence and machine learning for novel diabetes mellitus medications development

Diabetes mellitus, stemming from either insulin resistance or inadequate insulin secretion, represents a complex ailment that results in prolonged hyperglycemia and severe complications. Patients endure severe ramifications such as kidney disease, vision impairment, cardiovascular disorders, and susceptibility to infections, leading to significant physical suffering and imposing substantial socio-economic burdens. This condition has evolved into an increasingly severe health crisis. There is an urgent need to develop new treatments with improved efficacy and fewer adverse effects to meet clinical demands. However, novel drug development is costly, time-consuming, and often associated with side effects and suboptimal efficacy, making it a major challenge. Artificial Intelligence (AI) and Machine Learning (ML) have revolutionized drug development across its comprehensive lifecycle, spanning drug discovery, preclinical studies, clinical trials, and post-market surveillance. These technologies have significantly accelerated the identification of promising therapeutic candidates, optimized trial designs, and enhanced post-approval safety monitoring. Recent advances in AI, including data augmentation, interpretable AI, and integration of AI with traditional experimental methods, offer promising strategies for overcoming the challenges inherent in AI-based drug discovery. Despite these advancements, there exists a notable gap in comprehensive reviews detailing AI and ML applications throughout the entirety of developing medications for diabetes mellitus. This review aims to fill this gap by evaluating the impact and potential of AI and ML technologies at various stages of diabetes mellitus drug development. It does that by synthesizing current research findings and technological advances so as to effectively control diabetes mellitus and mitigate its far-reaching social and economic impacts. The integration of AI and ML promises to revolutionize diabetes mellitus treatment strategies, offering hope for improved patient outcomes and reduced healthcare burdens worldwide.

Use of Drug Sensitisers to Improve Therapeutic Index in Cancer

The clinical management of malignant tumours is challenging, often leading to severe adverse effects and death. Drug resistance (DR) antagonises the effectiveness of treatments, and increasing drug dosage can worsen the therapeutic index (TI). Current efforts to overcome DR predominantly involve the use of drug combinations, including applying multiple anti-cancerous drugs, employing drug sensitisers, which are chemical agents that enhance pharmacokinetics (PK), including the targeting of cellular pathways and regulating pertinent membrane transporters. While combining multiple compounds may lead to drug-drug interactions (DDI) or polypharmacy effect, the use of drug sensitisers permits rapid attainment of effective treatment dosages at the disease site to prevent early DR and minimise side effects and will reduce the chance of DDI as lower drug doses are required. This review highlights the essential use of TI in evaluating drug dosage for cancer treatment and discusses the lack of a unified standard for TI within the field. Commonly used benefit-risk assessment criteria are summarised, and the critical exploration of the current use of TI in the pharmaceutical industrial sector is included. Specifically, this review leads to the discussion of drug sensitisers to facilitate improved ratios of effective dose to toxic dose directly in humans. The combination of drug and sensitiser molecules might see additional benefits to rekindle those drugs that failed late-stage clinical trials by the removal of detrimental off-target activities through the use of lower drug doses. Drug combinations and employing drug sensitisers are potential means to combat DR. The evolution of drug combinations and polypharmacy on TI are reviewed. Notably, the novel binary weapon approach is introduced as a new opportunity to improve TI. This review emphasises the urgent need for a criterion to systematically evaluate drug safety and efficiency for practical implementation in the field.

A perfectly imperfect engine: Utilizing the digital twin paradigm in pulmonary hypertension

Pulmonary hypertension (PH) is a severe medical condition with a number of treatment options, the majority of which are introduced without consideration of the underlying mechanisms driving it within an individual and thus a lack of tailored approach to treatment. The one exception is a patient presenting with apparent pulmonary arterial hypertension and shown to have vaso-responsive disease, whose clinical course and prognosis is significantly improved by high dose calcium channel blockers. PH is however characterized by a relative abundance of available data from patient cohorts, ranging from molecular data characterizing gene and protein expression in different tissues to physiological data at the organ level and clinical information. Integrating available data with mechanistic information at the different scales into computational models suggests an approach to a more personalized treatment of the disease using model-based optimization of interventions for individual patients. That is, constructing digital twins of the disease, customized to a patient, promises to be a key technology for personalized medicine, with the aim of optimizing use of existing treatments and developing novel interventions, such as new drugs. This article presents a perspective on this approach in the context of a review of existing computational models for different aspects of the disease, and it lays out a roadmap for a path to realizing it.

Identifying patterning behavior in a plant infestation of insect pests

In this study, we developed a mechanistic model formulated as a system of reaction-diffusion equations (RDE) to explore the spatiotemporal dynamics of a theoretical pest with a tillering host plant in a controlled rectangular plant field. Local perturbation analysis, a recently developed method of analysis for wave propagation, was utilized to determine patterning regimes resulting from the local and global behaviors of the slow and fast diffusing components of the RDE system, respectively. Turing analysis was done to show that the RDE system does not exhibit Turing patterns. With bug mortality as the bifurcation parameter, regions with oscillations and stable coexistence of the pest and tillers were identified. Numerical simulations illustrate the patterning regimes in 1D and 2D settings. The oscillations suggest that recurrences in pest infestation is possible. Moreover, simulations showed that patterns produced in the model are strongly influenced by the pests' homogeneous dynamics inside the controlled environment.

Mathematical modeling of cancer immunotherapy for personalized clinical translation

Encouraging advances are being made in cancer immunotherapy modeling, especially in the key areas of developing personalized treatment strategies based on individual patient parameters, predicting treatment outcomes and optimizing immunotherapy synergy when used in combination with other treatment approaches. Here we present a focused review of the most recent mathematical modeling work on cancer immunotherapy with a focus on clinical translatability. It can be seen that this field is transitioning from pure basic science to applications that can make impactful differences in patients' lives. We discuss how researchers are integrating experimental and clinical data to fully inform models so that they can be applied for clinical predictions, and present the challenges that remain to be overcome if widespread clinical adaptation is to be realized. Lastly, we discuss the most promising future applications and areas that are expected to be the focus of extensive upcoming modeling studies.

Model comparison via simplicial complexes and persistent homology

In many scientific and technological contexts, we have only a poor understanding of the structure and details of appropriate mathematical models. We often, therefore, need to compare different models. With available data we can use formal statistical model selection to compare and contrast the ability of different mathematical models to describe such data. There is, however, a lack of rigorous methods to compare different models a priori. Here, we develop and illustrate two such approaches that allow us to compare model structures in a systematic way by representing models as simplicial complexes. Using well-developed concepts from simplicial algebraic topology, we define a distance between models based on their simplicial representations. Employing persistent homology with a flat filtration provides for alternative representations of the models as persistence intervals, which represent model structure, from which the model distances are also obtained. We then expand on this measure of model distance to study the concept of model equivalence to determine the conceptual similarity of models. We apply our methodology for model comparison to demonstrate an equivalence between a positional-information model and a Turing-pattern model from developmental biology, constituting a novel observation for two classes of models that were previously regarded as unrelated.

A Vision of Future Healthcare: Potential Opportunities and Risks of Systems Medicine from a Citizen and Patient Perspective-Results of a Qualitative Study

Advances in (bio)medicine and technological innovations make it possible to combine high-dimensional, heterogeneous health data to better understand causes of diseases and make them usable for predictive, preventive, and precision medicine. This study aimed to determine views on and expectations of “systems medicine” from the perspective of citizens and patients in six focus group interviews, all transcribed verbatim and content analyzed. A future vision of the use of systems medicine in healthcare served as a stimulus for the discussion. The results show that although certain aspects of systems medicine were seen positive (e.g., use of smart technology, digitalization, and networking in healthcare), the perceived risks dominated. The high degree of technification was perceived as emotionally burdensome (e.g., reduction of people to their data, loss of control, dehumanization). The risk-benefit balance for the use of risk-prediction models for disease events and trajectories was rated as rather negative. There were normative and ethical concerns about unwanted data use, discrimination, and restriction of fundamental rights. These concerns and needs of citizens and patients must be addressed in policy frameworks and health policy implementation strategies to reduce negative emotions and attitudes toward systems medicine and to take advantage of its opportunities.

IMFLer: A Web Application for Interactive Metabolic Flux Analysis and Visualization

Increasing genome-wide data in biological sciences and medicine has contributed to the development of a variety of visualization tools. Several automatic, semiautomatic, and manual visualization tools have already been developed. Some even have integrated flux balance analysis (FBA), but in most cases, it depends on separately installed third party software that is proprietary and does not allow customization of its functionality and has many restrictions for easy data distribution and analysis. In this study, we present an interactive metabolic flux analyzer and visualizer (IMFLer)-a static single-page web application that enables the reading and management of metabolic model layout maps, as well as immediate visualization of results from both FBA and flux variability analysis (FVA). IMFLer uses the Escher Builder tool to load, show, edit, and save metabolic pathway maps. This makes IMFLer an attractive and easily applicable tool with a user-friendly interface. Moreover, it allows to faster interpret results from FBA and FVA and improves data interoperability by using a standardized file format for the genome-scale metabolic model. IMFLer is a fully open-source tool that enables the rapid visualization and interpretation of the results of FBA and FVA with no time setup and no programming skills required, available at https://lv-csbg.github.io/IMFLer/.

Metformin Transport Rates Between Plasma and Red Blood Cells in Humans

Background

Metformin has been used for the treatment of type 2 diabetes for over 60 years; however, its mechanism of pharmacological action is not fully clear. Different hypotheses exist regarding metformin distribution and redistribution mechanisms between plasma and erythrocytes/red blood cells (RBCs).

Objective

We aimed to test the hypothesis that the metformin distribution between plasma and RBC occurs via concentration difference-driven passive transport and estimated transport rate coefficient values based on metformin concentration time series in plasma and RBCs from in vivo studies.

Methods

An ordinary differential equation (ODE) system with two compartments was used to describe diffusion-based passive transport between plasma and RBCs. Metformin concentration time series in plasma and RBCs of 35 individuals were used for metformin transport parametrization. Plasma concentration has been approximated by biexponential decline.

Results

A single passive transport coefficient, k = 0.044 ± 0.014 (h^–1), can be applied, describing the uptake and release transport rate versus the linear equation v = k × (M_pl − M_RBC), where M_pl is the metformin concentration in plasma and M_RBC is the metformin concentration in RBCs.

Conclusions

Our research suggests that passive transport can explain metformin distribution dynamics between plasma and RBCs because transport speed is proportional to the metformin concentration difference and independent of the transport direction. Concentration difference-driven passive transport can explain the mechanism of faster metformin distribution to RBCs the first few hours after administration, and faster release and domination of the redistribution transport rate after metformin concentration in plasma becomes smaller than in RBCs.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40262-021-01058-2.

Physiologically based metformin pharmacokinetics model of mice and scale-up to humans for the estimation of concentrations in various tissues

Metformin is the primary drug for type 2 diabetes treatment and a promising candidate for other disease treatment. It has significant deviations between individuals in therapy efficiency and pharmacokinetics, leading to the administration of an unnecessary overdose or an insufficient dose. There is a lack of data regarding the concentration-time profiles in various human tissues that limits the understanding of pharmacokinetics and hinders the development of precision therapies for individual patients. The physiologically based pharmacokinetic (PBPK) model developed in this study is based on humans’ known physiological parameters (blood flow, tissue volume, and others). The missing tissue-specific pharmacokinetics parameters are estimated by developing a PBPK model of metformin in mice where the concentration time series in various tissues have been measured. Some parameters are adapted from human intestine cell culture experiments. The resulting PBPK model for metformin in humans includes 21 tissues and body fluids compartments and can simulate metformin concentration in the stomach, small intestine, liver, kidney, heart, skeletal muscle adipose, and brain depending on the body weight, dose, and administration regimen. Simulations for humans with a bodyweight of 70kg have been analyzed for doses in the range of 500-1500mg. Most tissues have a half-life (T_1/2) similar to plasma (3.7h) except for the liver and intestine with shorter T_1/2 and muscle, kidney, and red blood cells that have longer T_1/2. The highest maximal concentrations (C_max) turned out to be in the intestine (absorption process) and kidney (excretion process), followed by the liver. The developed metformin PBPK model for mice does not have a compartment for red blood cells and consists of 20 compartments. The developed human model can be personalized by adapting measurable values (tissue volumes, blood flow) and measuring metformin concentration time-course in blood and urine after a single dose of metformin. The personalized model can be used as a decision support tool for precision therapy development for individuals.

Driving success in personalized medicine through AI-enabled computational modeling

The development of successful drugs is expensive and time-consuming because of high clinical attrition rates. This is caused partially by the rupture seen in the translatability of the drug from the bench to the clinic in the context of personalized medicine. Artificial intelligence (AI)-driven platforms integrated with mechanistic modeling have become instrumental in accelerating the drug development process by leveraging data ubiquitously across the various phases. AI can counter the deficiencies and ambiguities that arise during the classical drug development process while reducing human intervention and bridging the translational gap in discovering the connections between drugs and diseases.